[Linux-cluster] CentOS5.x86_64 HA Cluster will not fail over to remaining node

[Lon Hohberger]

On Thu, 2007-10-11 at 08:40 -0600, Christian Ullman wrote:

> nodes: ccweb1 & ccweb2, which is an httpd cluster only
> 
> NOTE: no shared storages in employed as the content is static and the
> all the dynamic data is stored on another database cluster. 
> 
>  
> 
> When the service "ca1-ccweb", which consists of a floating IP
> and /etc/init.d/httpd script (see config file below), running on a
> ccweb1 node and is fenced via:
> 
>  
> 
> fence_ipmilan -P -v -i 10.1.2.13 -l ADMIN -p password -A password -o
> off


> "clustat" reports the service is stopped but does not restart it on
> the remaining node:

It's in the 'stopping' state.  It looks like you hit this, but it's been
fixed for some time; I don't see why you'd hit that on any RHEL5
package:

https://bugzilla.redhat.com/show_bug.cgi?id=193255

If the service is in the stopping state, failover should work.  It's
almost like rgmanager's waiting for the node to be fenced (which should
have already happened as you said).

What rgmanager package do you have (rpm -q rgmanager)?

-- Lon